OdetteDuLac Contributions

Idea Formulation Data Discovery Setup and Data Download Rationale and Research Questions Data Description (Dataset Information) Exploratory Analysis README

Setup and Data

## Reading layer `geoBoundaries-TUR-ADM0_simplified' from data source 
##   `/Users/sbamba/Documents/Professional/Github/wproj/DistributionCharacteristics2023TurkyeEarthquake/Data/Shapefiles/geoBoundaries-TUR-ADM0-all/geoBoundaries-TUR-ADM0_simplified.shp' 
##   using driver `ESRI Shapefile'
## Simple feature collection with 1 feature and 5 fields
## Geometry type: MULTIPOLYGON
## Dimension:     XY
## Bounding box:  xmin: 25.66545 ymin: 35.80768 xmax: 44.81766 ymax: 42.1048
## Geodetic CRS:  WGS 84
## Reading layer `geoBoundaries-TUR-ADM1_simplified' from data source 
##   `/Users/sbamba/Documents/Professional/Github/wproj/DistributionCharacteristics2023TurkyeEarthquake/Data/Shapefiles/geoBoundaries-TUR-ADM1-all/geoBoundaries-TUR-ADM1_simplified.shp' 
##   using driver `ESRI Shapefile'
## Simple feature collection with 81 features and 5 fields
## Geometry type: MULTIPOLYGON
## Dimension:     XY
## Bounding box:  xmin: 25.66545 ymin: 35.80768 xmax: 44.81766 ymax: 42.1048
## Geodetic CRS:  WGS 84
## Reading layer `geoBoundaries-TUR-ADM2_simplified' from data source 
##   `/Users/sbamba/Documents/Professional/Github/wproj/DistributionCharacteristics2023TurkyeEarthquake/Data/Shapefiles/geoBoundaries-TUR-ADM2-all/geoBoundaries-TUR-ADM2_simplified.shp' 
##   using driver `ESRI Shapefile'
## Simple feature collection with 973 features and 5 fields
## Geometry type: MULTIPOLYGON
## Dimension:     XY
## Bounding box:  xmin: 25.66545 ymin: 35.80805 xmax: 44.81766 ymax: 42.1048
## Geodetic CRS:  WGS 84

Rationale and Research Questions

On February 6, 2023 a 7.8 magnitude earthquake hit northwest section Gaziantep, it’s epicenter near the town of Nurdağı.Less than 12 hours later another 7.5 magnitude earthquake the neighboring province of Kahramanmaraş. The earthquakes had severe damaging impacts in 11 provinces of southeast Türkiye as well as northwest Syria, including in the cities of Idlib and Aleppo, and their effects were felt as far away as Lebanon, Palestine, and Egypt. Strong aftershocks from the earthquakes lasted for more than a month (Kawoosa, 2023), leading to the continued destruction of buildings already damaged in the initial earthquakes. Some aftershocks are still occurring today.

The earthquakes caused more than 55,000 people were killed and more than 130,000 people were injured across Türkiye and Syria as a result of the earthquakes, and tens of thousands of people are missing and millions displaced.

To learn more about some of the impacts of the earthquake we decided to examine the distribution and characteristics of 2023 Türkiye-Syria earthquakes and their aftershocks within the borders of Türkiye to understand the relationship between earthquake magnitude and depth using GLMs and visually compares the occurrence of building destruction with these earthquake variables as well as the human response to the destruction, as represented by twitter activity.

To conduct this analysis we used data from the USGS Earthquake Catalog on the earthquakes and data from the UN Office of Humanitarian Affairs’ Humanitarian Data Exchange portal on building damage and twitter activity. We also used administrative boundary data from the geoBoundaries package created by the College of William and Mary’s geoLab.

Research Questions:

  1. How is earthquake depth correlated to earthquake magnitude?
  2. How are depth and magnitude related to destruction of buildings?
  3. What is the distribution of earthquake-related twitter activity?
  4. How does this distribution look in relation to the distribution of destroyed buildings?

Dataset Information

This project used four main datasets: Administrative Boundary Layers, Earthquake Data, Destroyed Buildings Data, and Twitter Activity Data.

The Administrative Boundary Layers were downloaded from the geoBoundaries dataset created by the geoLab at the College of William and Mary. Administrative boundaries for the country, province, and district level were downloaded and saved as shapefiles.

The Earthquake Data was downloaded from the US Geological Survey Earthquake Catalog using their API. It is saved in a csv format and is a point layer of earthquake data from February 5 to march 5, 2023. The dataset includes measures of depth, magnitude, time, latitude, and longitude as well as other statistics related to measuring and detection errors.

The Destroyed Buildings Data was pulled from the UN Office of UN Office of Humanitarian Affairs’ Humanitarian Data Exchange Portal. It is saved in a csv format and is a point layer of destroyed building data that is updated daily and curated using Open Street Map and Map Roulette.

The Twitter Activity Data was pulled from the UN Office of UN Office of Humanitarian Affairs’ Humanitarian Data Exchange Portal. It is saved in a xlsx format and is tabular data aggregated from points to administrative level 2 (distric) polygons. It includes separate files of missing people reports, help requests, shelter requests, and damage reports shared through images.

Because of the very specific nature of the datasets and the spatial analysis conducted, the data did not require serious wrangling. Some datasets, such as the Earthquake dataset, were wrangled and processed to create new data and time columns using lubridate and distance columns. Some datasets, such as the administrative boundary layers and the Earthquake dataset, were reprojected to allow for spatial analayes, such as interpolation.

Table 1 shows a summary of the dataset information.

Table 1: Summary of Datasets
Name Description Variables Information Original Projection Format Source
Administrative Boundary Layers polygons name; type (ADM level); geometry ADM 0, ADM 1, and ADM 2 4326 shapefile geoBoundaries
Earthquake Data points date/time; latitude; longitude; depth (km); magnitude; place; measures of error; measurement sources; ID earthquakes Feb. 5-Mar. 5, 2023 4326 .csv USGS Earthquake Catalog
Destroyed Buildings Data points source; damage_event; date; longigute; latitude destroyed buildings 4326 .csv UN OCHA HDX
Twitter Activity Data points aggregated to ADM 2 polygons district; population; frequency of images; frequency of tweets; impact score; impact category help requests, shelter requests, missing people reports, damage reports (images) NA .xlsx UN OCHA HDX

Exploratory Analysis

The map below shows extent of Turkiye within which our analysis of the earthquake related data will take place. Moreover, the administrative boundaries for provinces and districts are shown with their labels. This will aid in situating the earthquake, damage, and twitter data within the affected 11 provinces and their relevant districts.

Figure 1 below illustrates the rough distribution of the earthquakes that occurred between February 5 and March 5, 2023. The earthquakes occurred in a large south to north swath from Hatay to the Black Sea region, with most of the earthquakes concentrating in the South, particularly near Hatay, Gaziantep, Şanlıurfa, and Kahramanmaraş.

##                       time latitude longitude  depth mag magType nst gap  dmin
## 1 2023-03-04T05:27:34.779Z  38.0299   38.2134  2.860 4.2      mb  35  82 0.760
## 2 2023-03-03T02:53:43.065Z  37.8395   36.7004  7.120 5.0     mww  85  26 0.369
## 3 2023-03-02T13:35:54.000Z  38.2315   38.1119  6.914 3.4      ml  16  70 0.878
## 4 2023-03-01T16:37:59.849Z  37.2659   37.0676  9.357 4.4      mb  39 104 0.147
## 5 2023-03-01T07:20:23.043Z  36.3065   36.0970 10.000 4.7      mb  87  66 1.243
## 6 2023-03-01T04:15:31.399Z  37.7611   37.8869 10.000 4.3      mb  31 109 0.798
##    rms net         id                  updated                          place
## 1 0.71  us us7000jh8x 2023-05-08T02:00:43.040Z   2 km WNW of Çelikhan, Turkey
## 2 0.90  us us7000jgyv 2023-05-08T02:00:31.040Z     26 km SE of Göksun, Turkey
## 3 0.71  us us7000jgt5 2023-05-08T02:00:20.040Z 13 km WSW of Ye?ilyurt, Turkey
## 4 0.78  us us7000jgly 2023-05-08T02:00:10.040Z   31 km ENE of Nurda??, Turkey
## 5 0.92  us us7000jgic 2023-05-08T02:00:06.040Z      7 km W of Anayazi, Turkey
## 6 0.74  us us7000jgev 2023-05-08T02:00:03.040Z         4 km SW of Tut, Turkey
##         type horizontalError depthError magError magNst   status locationSource
## 1 earthquake            4.71      5.136    0.147     13 reviewed             us
## 2 earthquake            5.78      3.331    0.059     28 reviewed             us
## 3 earthquake            1.65      7.730    0.069     28 reviewed             us
## 4 earthquake            5.27      2.865    0.138     15 reviewed             us
## 5 earthquake            5.57      1.840    0.072     58 reviewed             us
## 6 earthquake            2.28      1.952    0.218      6 reviewed             us
##   magSource
## 1        us
## 2        us
## 3        us
## 4        us
## 5        us
## 6        us
Distirbution of Earthquake Points

Figure 1: Distirbution of Earthquake Points

Analysis

Running a Generalized Linear regression to examine the relationship between Depth and Magnitude.

After running a simple single linear regression, there is very little relationship between depth and magnitude of earthquakes, the r-squared is 0.0006317. This suggests that the variation in depth cannot be explained by magnitude.

## 
## Call:
## lm(formula = EQC.data.df$mag ~ EQC.data.df$depth)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.0508 -0.2471 -0.0523  0.1261  3.3529 
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        4.459026   0.067333  66.224   <2e-16 ***
## EQC.data.df$depth -0.001189   0.006014  -0.198    0.843    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.42 on 456 degrees of freedom
## Multiple R-squared:  8.576e-05,  Adjusted R-squared:  -0.002107 
## F-statistic: 0.03911 on 1 and 456 DF,  p-value: 0.8433
## 
## Call:
## lm(formula = idw.rast.mag ~ idw.rast.depth, data = v)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.31738 -0.00266  0.00284  0.00391  1.08610 
## 
## Coefficients:
##                  Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     4.4918701  0.0090163 498.196  < 2e-16 ***
## idw.rast.depth -0.0045394  0.0008444  -5.376 7.83e-08 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.02786 on 7924 degrees of freedom
## Multiple R-squared:  0.003634,   Adjusted R-squared:  0.003508 
## F-statistic:  28.9 on 1 and 7924 DF,  p-value: 7.834e-08

Running a Generalized Linear regression to examine the relationship between Depth of aftershocks and distance from mainshock.

Running the same generalized linear model, there is no relationship between depth and distance from the mainshock with an r-squared value of -0.0001214. The variance in depth cannot be explained by distance from the main shock. In the figure 3 below there is a clear cluster of aftershocks at 10km of depth, this is the ‘fixed earthquake depth’ which we dicuss more in the Conclusions Section.

## 
## Call:
## lm(formula = depth ~ distance[, 1], data = EQC.data.df)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -10.3393  -0.8155  -0.6787   0.7444  16.0309 
## 
## Coefficients:
##                Estimate Std. Error t value Pr(>|t|)    
## (Intercept)   10.473364   0.304261  34.422   <2e-16 ***
## distance[, 1]  0.002166   0.002411   0.898     0.37    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.267 on 456 degrees of freedom
## Multiple R-squared:  0.001766,   Adjusted R-squared:  -0.0004228 
## F-statistic: 0.8068 on 1 and 456 DF,  p-value: 0.3695
Generalized Linear Model relating Depth of aftershocks and Distance from mainshock

Figure 3: Generalized Linear Model relating Depth of aftershocks and Distance from mainshock

Running a Generalized Linear regression to examine the relationship between Magnitude of aftershocks and distance from mainshock.

Our analysis also indicated no relationship between magnitude and distance from the mainshock with an r-squared value of 0.0007319.You can see from the figure 4 that there is a clustering of aftershocks 0-200 km away between 4 and 5 magnitude. However, as distance increases there is no correlation to magnitude.

## 
## Call:
## lm(formula = mag ~ distance[, 1], data = EQC.data.df)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.0325 -0.2489 -0.0620  0.1224  3.3147 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)    4.4853413  0.0390538 114.850   <2e-16 ***
## distance[, 1] -0.0003578  0.0003095  -1.156    0.248    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.4194 on 456 degrees of freedom
## Multiple R-squared:  0.002922,   Adjusted R-squared:  0.0007357 
## F-statistic: 1.336 on 1 and 456 DF,  p-value: 0.2483
Generalized Linear Model relating Magnitude of aftershocks and Distance from mainshock

Figure 4: Generalized Linear Model relating Magnitude of aftershocks and Distance from mainshock

##Running a Generalized Linear regression to examine the relationship between Depth of aftershocks and days since mainshock.

Our most significant relationship was the only multiple linear regression test we ran. Looking at both date and time, the variance in depth was 3.771% explained, with a p-value of 0.00005982. This was primarily due to date, with a single linear regression r-squared value of 0.03278. As can be seen in figure 5 many more aftershocks occur in the first 24 hours than later, so the relationship could be partially explained by this tapering off in frequency.

## 
## Call:
## lm(formula = depth ~ date, data = EQC.data.df)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -9.8746 -1.1095 -0.8865  1.1131 15.7950 
## 
## Coefficients:
##               Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 2173.26167  491.08301   4.425 1.21e-05 ***
## date          -0.11149    0.02532  -4.404 1.33e-05 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.203 on 456 degrees of freedom
## Multiple R-squared:  0.04079,    Adjusted R-squared:  0.03869 
## F-statistic: 19.39 on 1 and 456 DF,  p-value: 1.328e-05
## 
## Call:
## lm(formula = depth ~ hms, data = EQC.data.df)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -10.1536  -1.0435  -0.5894   0.7247  16.3941 
## 
## Coefficients:
##               Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  1.114e+01  2.878e-01  38.722   <2e-16 ***
## hms         -1.134e-05  6.385e-06  -1.776   0.0763 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.259 on 456 degrees of freedom
## Multiple R-squared:  0.006873,   Adjusted R-squared:  0.004695 
## F-statistic: 3.156 on 1 and 456 DF,  p-value: 0.07633
## 
## Call:
## lm(formula = depth ~ date + hms, data = EQC.data.df)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -9.6742 -1.3043 -0.7762  1.0539 16.0944 
## 
## Coefficients:
##               Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  2.158e+03  4.901e+02   4.404 1.33e-05 ***
## date        -1.107e-01  2.527e-02  -4.381 1.47e-05 ***
## hms         -1.085e-05  6.262e-06  -1.732   0.0839 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 3.196 on 455 degrees of freedom
## Multiple R-squared:  0.04708,    Adjusted R-squared:  0.04289 
## F-statistic: 11.24 on 2 and 455 DF,  p-value: 1.721e-05
Generalized Linear Model relating Depth of aftershocks and Time from mainshock

Figure 5: Generalized Linear Model relating Depth of aftershocks and Time from mainshock

Running a Generalized Linear regression to examine the relationship between Magnitude of aftershocks and days since mainshock.

Magnitude was 2.45% explained by the multiple linear regression with date and time. This relationship was significant with a p-value of 0.001321. As can be seen above and in figure 6 many more aftershocks occur in the first 24 hours than later, so the relationship could be partially explained by this tapering off in frequency.

## 
## Call:
## lm(formula = mag ~ date, data = EQC.data.df)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -0.8185 -0.2500 -0.0863  0.1137  3.3137 
## 
## Coefficients:
##               Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 220.908119  63.594459   3.474 0.000562 ***
## date         -0.011159   0.003278  -3.404 0.000723 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.4148 on 456 degrees of freedom
## Multiple R-squared:  0.02478,    Adjusted R-squared:  0.02264 
## F-statistic: 11.59 on 1 and 456 DF,  p-value: 0.0007233
## 
## Call:
## lm(formula = mag ~ hms, data = EQC.data.df)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -1.0328 -0.2438 -0.0848  0.1108  3.3115 
## 
## Coefficients:
##               Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  4.494e+00  3.699e-02 121.493   <2e-16 ***
## hms         -1.257e-06  8.208e-07  -1.532    0.126    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.4189 on 456 degrees of freedom
## Multiple R-squared:  0.005118,   Adjusted R-squared:  0.002936 
## F-statistic: 2.346 on 1 and 456 DF,  p-value: 0.1263
## 
## Call:
## lm(formula = mag ~ date + hms, data = EQC.data.df)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -0.8073 -0.2245 -0.0929  0.1003  3.2735 
## 
## Coefficients:
##               Estimate Std. Error t value Pr(>|t|)    
## (Intercept)  2.193e+02  6.352e+01   3.452 0.000609 ***
## date        -1.107e-02  3.275e-03  -3.381 0.000785 ***
## hms         -1.208e-06  8.117e-07  -1.488 0.137476    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.4142 on 455 degrees of freedom
## Multiple R-squared:  0.0295, Adjusted R-squared:  0.02523 
## F-statistic: 6.915 on 2 and 455 DF,  p-value: 0.0011
Generalized Linear Model relating Magnitude of aftershocks and Time from mainshock

Figure 6: Generalized Linear Model relating Magnitude of aftershocks and Time from mainshock

Help Requests

Help requests, 7, surrounding the two >7 magnitude earthquake points range from Low, Medium, High, and Very High impact scores. However, the “Very High” impact scores exist nearest these points and rarely deviate from them (with few exception boundaries). The density of destroyed buildings is hardly reflected by the number of help requests in those same areas, 2.

## List of 3
##  $ legend.text    :List of 11
##   ..$ family       : NULL
##   ..$ face         : NULL
##   ..$ colour       : NULL
##   ..$ size         : num 12
##   ..$ hjust        : NULL
##   ..$ vjust        : NULL
##   ..$ angle        : NULL
##   ..$ lineheight   : NULL
##   ..$ margin       : NULL
##   ..$ debug        : NULL
##   ..$ inherit.blank: logi FALSE
##   ..- attr(*, "class")= chr [1:2] "element_text" "element"
##  $ legend.title   :List of 11
##   ..$ family       : NULL
##   ..$ face         : NULL
##   ..$ colour       : NULL
##   ..$ size         : num 12
##   ..$ hjust        : NULL
##   ..$ vjust        : NULL
##   ..$ angle        : NULL
##   ..$ lineheight   : NULL
##   ..$ margin       : NULL
##   ..$ debug        : NULL
##   ..$ inherit.blank: logi FALSE
##   ..- attr(*, "class")= chr [1:2] "element_text" "element"
##  $ legend.position: chr "bottom"
##  - attr(*, "class")= chr [1:2] "theme" "gg"
##  - attr(*, "complete")= logi FALSE
##  - attr(*, "validate")= logi TRUE
Distribution of Help Requests

Figure 7: Distribution of Help Requests

Shelter Requests

Most boundaries containing “High” and “Very High” impact categories for requesting shelter, 8, are located in the boundaries containing the most damaged/destroyed buildings. Both High and Very High categories exist on or surrounding the >7 magnitude earthquake points. “Low” to “No Reports” exist further from the >7 magnitude points.

## List of 3
##  $ legend.text    :List of 11
##   ..$ family       : NULL
##   ..$ face         : NULL
##   ..$ colour       : NULL
##   ..$ size         : num 12
##   ..$ hjust        : NULL
##   ..$ vjust        : NULL
##   ..$ angle        : NULL
##   ..$ lineheight   : NULL
##   ..$ margin       : NULL
##   ..$ debug        : NULL
##   ..$ inherit.blank: logi FALSE
##   ..- attr(*, "class")= chr [1:2] "element_text" "element"
##  $ legend.title   :List of 11
##   ..$ family       : NULL
##   ..$ face         : NULL
##   ..$ colour       : NULL
##   ..$ size         : num 12
##   ..$ hjust        : NULL
##   ..$ vjust        : NULL
##   ..$ angle        : NULL
##   ..$ lineheight   : NULL
##   ..$ margin       : NULL
##   ..$ debug        : NULL
##   ..$ inherit.blank: logi FALSE
##   ..- attr(*, "class")= chr [1:2] "element_text" "element"
##  $ legend.position: chr "bottom"
##  - attr(*, "class")= chr [1:2] "theme" "gg"
##  - attr(*, "complete")= logi FALSE
##  - attr(*, "validate")= logi TRUE
Distirbution of Shelter Requests

Figure 8: Distirbution of Shelter Requests

Missing persons

Our visualization plot for missing persons reports, 9, was most significant in containing the “No Reports” impact category in comparison to all of our Twitter data plots. Where the most No Reports exist, these boundaries do not contain the density (nor any) damaged/destroyed building points.

## List of 3
##  $ legend.text    :List of 11
##   ..$ family       : NULL
##   ..$ face         : NULL
##   ..$ colour       : NULL
##   ..$ size         : num 12
##   ..$ hjust        : NULL
##   ..$ vjust        : NULL
##   ..$ angle        : NULL
##   ..$ lineheight   : NULL
##   ..$ margin       : NULL
##   ..$ debug        : NULL
##   ..$ inherit.blank: logi FALSE
##   ..- attr(*, "class")= chr [1:2] "element_text" "element"
##  $ legend.title   :List of 11
##   ..$ family       : NULL
##   ..$ face         : NULL
##   ..$ colour       : NULL
##   ..$ size         : num 12
##   ..$ hjust        : NULL
##   ..$ vjust        : NULL
##   ..$ angle        : NULL
##   ..$ lineheight   : NULL
##   ..$ margin       : NULL
##   ..$ debug        : NULL
##   ..$ inherit.blank: logi FALSE
##   ..- attr(*, "class")= chr [1:2] "element_text" "element"
##  $ legend.position: chr "bottom"
##  - attr(*, "class")= chr [1:2] "theme" "gg"
##  - attr(*, "complete")= logi FALSE
##  - attr(*, "validate")= logi TRUE
Distirbution of Missing Persons Reports

Figure 9: Distirbution of Missing Persons Reports

Reported Damage

All reported damage Very High, High, and Medium category points, 10, exist where damaged/destroyed buildings were reported. In comparison, all Low reported damage exist where there are little to no reported destroyed/damaged buildings, 2.

## List of 3
##  $ legend.text    :List of 11
##   ..$ family       : NULL
##   ..$ face         : NULL
##   ..$ colour       : NULL
##   ..$ size         : num 12
##   ..$ hjust        : NULL
##   ..$ vjust        : NULL
##   ..$ angle        : NULL
##   ..$ lineheight   : NULL
##   ..$ margin       : NULL
##   ..$ debug        : NULL
##   ..$ inherit.blank: logi FALSE
##   ..- attr(*, "class")= chr [1:2] "element_text" "element"
##  $ legend.title   :List of 11
##   ..$ family       : NULL
##   ..$ face         : NULL
##   ..$ colour       : NULL
##   ..$ size         : num 12
##   ..$ hjust        : NULL
##   ..$ vjust        : NULL
##   ..$ angle        : NULL
##   ..$ lineheight   : NULL
##   ..$ margin       : NULL
##   ..$ debug        : NULL
##   ..$ inherit.blank: logi FALSE
##   ..- attr(*, "class")= chr [1:2] "element_text" "element"
##  $ legend.position: chr "bottom"
##  - attr(*, "class")= chr [1:2] "theme" "gg"
##  - attr(*, "complete")= logi FALSE
##  - attr(*, "validate")= logi TRUE
Distirbution of Reported Damage

Figure 10: Distirbution of Reported Damage

Summary and Conclusions

Below we get into the statistical analyses we did, looking at magnitude and depth related to each as well as distance and time from the 7.8 magnitude mainshock. Before getting into that, a quick explanation of how earthquakes are measured. Magnitude is the strength of the earthquake, measured directly where the rock breaks or shifts and is related to ‘strain-drop’ and how much the rock is moving after the break or shift that occurred in the Earth’s crust after it built up too much tension (or strain) and the size of the seismic waves that will travel through the crust. In other cases intensity is of more interest, the USGS uses the Moderated Mercalli Scale. This is a measure of how people experience the earthquake, with weak (“Many people don’t recognize it as an earthquake, vibrations are similar to a passing truck”) to extreme (“Some well-build structures destroyed, most masonry and frame structures with foundations destroyed, rails bent”). Intensity is related to both magnitude and depth; higher magnitude earthquakes occurring at shallower depth will result in higher intensity. The earthquake we are looking at was a 7.8 magnitude earthquake that occurred at 10 km depth.

The statistical result of no relationship between magnitude and depth is backed up by the research on the subject. Earthquakes occur in the crust or upper mantle of the Earth, ranging from 0km to 800 km in depth. Most earthquakes are ‘shallow’ occurring at depths of 50km or less because the rock is colder and more brittle at this depth and so more likely to break under the strain and deformation of the plate. Deeper earthquakes are more likely to occur at subducting plate boundaries, such as the converging zone in the Pacific Ocean.

The variance in depth cannot be explained by distance from the main shock. Research suggests that ‘shallow’ earthquakes produce more aftershocks. You can see in the data that most of the earthquakes seem to occur at 10km of depth. According the the USGS 10km is a ‘fixed earthquake depth’, when earthquake data can’t be used to compute a reliable depth 10km is used. In most earthquake prone zones of the world, reliable depths tend to average 10km or close to it. This may explain the cluster of earthquakes at 10km.

This was another instance of the variance in magnitude(or depth) being less than 1% explained by our independent variable, in this case distance from the mainshock. Aftershocks are smaller earthquakes that occur on the same fault, as the rock ‘realigns’ after the ‘strain drop’ or the large shift that occurred and caused the mainshock of the first earthquake.

Our most significant relationship was the only multiple linear regression test we ran for both magnitude and depth. This was primarily due to date in both cases. This relationship is possibly due to the global trend of earthquakes occurring at 10km as well as the drop off in frequency of aftershocks after the mainshock. In the case of magnitude it is probably more due to the decrease in frequency after the first 24 hours. At first this result was somewhat surprising, however, after research according to USGS aftershocks typically taper off in frequency but not magnitude after the mainshock, with strong earthquakes possible late in the series. Aftershocks shortly after the mainshock can be just as large or larger, as with the case of our 7.5 magnitude earthquake occurring less than 24 hours after the mainshock of 7.8 magnitude. This secondary large earthquake caused its own series of aftershocks. Other notable aftershocks include a 6.4 magnitude earthquake on February 20th and a 5.6 magnitude earthquake on February 27th. Aftershocks decrease in frequency within a few days after the mainshock but can still last for weeks to years, with each larger aftershock triggering its own series of aftershocks. This is the case in Türkiye and Syria with aftershocks still occurring today.

References

Administrative boundaries courtesy of geoBoundaries

“Aftershock Forecast Overview.” Accessed May 1, 2023. https://earthquake.usgs.gov/data/oaf/overview.php.

“At What Depth Do Earthquakes Occur? What Is the Significance of the Depth? | U.S. Geological Survey.” Accessed May 1, 2023. https://www.usgs.gov/faqs/what-depth-do-earthquakes-occur-what-significance-depth.

“Earthquakes: What? Where? How Are They Measured? – Deep Geothermal Heat Research.” Accessed May 1, 2023. https://deepgeothermalheat.engineering.cornell.edu/cubo-science-intro/earthquakes-what-where-how-are-they-measured/.

https://quantectum.com/. “Facts About Earthquake’s Depth.” Accessed May 1, 2023. https://quantectum.com/blog/facts-about-earthquakes-depth/.

Kawoosa, V.M. (2023, March 1). How Turkey has been inundated with aftershocks. Reuters. https://www.reuters.com/graphics/TURKEY-QUAKE/AFTERSHOCKS/dwpkdzklevm/

Vallée, Martin. “Source Time Function Properties Indicate a Strain Drop Independent of Earthquake Depth and Magnitude.” Nature Communications 4, no. 1 (October 15, 2013): 2606. https://doi.org/10.1038/ncomms3606.

Webb, Paul. “4.8 Earthquakes and Plate Tectonics.” Accessed May 1, 2023. https://rwu.pressbooks.pub/webboceanography/chapter/4-8-earthquakes-and-plate-tectonics/.

“What Is the Difference between Aftershocks and Swarms? | U.S. Geological Survey.” Accessed May 1, 2023. https://www.usgs.gov/faqs/what-difference-between-aftershocks-and-swarms.